Is a picture worth 1000 votes? Analyzing the sentiment of election related social photos

This paper explores techniques for automatically recognizing the sentiment of facial expressions in social photos, especially those of politicians in the context of elections. We first use the Active Shape Model (ASM) to extract facial feature points. Next, the shape model points from the ASM are normalized to a standard shape and then submitted to a trained AdaBoost classifier to recognize the sentiment of facial expressions. Three types of sentiment are of primary interest: flattering, neutral and unflattering. Finally, the approach is evaluated by experiments, which indicate that the proposed method is sufficiently effective for facial expression analysis of images of election candidates and thus can be used to gauge the public opinion during the election.